*=equal contribution, #=corresponding author

A human-to-robot learning that bridges the embodiment gap with transitive 2D feature distillation and explicit 3D geometry alignment.

A dual-expert Vision-Language-Action framework that injects one-step partially denoised 3D scene flow as a latent motion prior.

A cross-embodiment framework that transfers wheeled-humanoid data to bipedal VLA models via morphology-agnostic 6D end-effector trajectories and a heuristic-enhanced online DAgger controller.

POEM-v2 a generalizable multi-view 3D hand reconstruction model trained on large-scale multi-view datasets. It enables accurate, flexible, and occlusion-robust hand mesh recovery across arbitrary multi-view setups.

A low-cost exoskeleton system for large-scale in-the-wild demonstration collection. It transforms the collected in-the-wild demonstrations into pseudo-robot demonstrations.

A bidirectional robotic autoregressive policy, which infers trajectories by gradually expanding actions from sparse keyframes, demonstrated exceeding diffusion policies.

A two cascaded diffusion processes for object motion generation and robot action generation under object motion guidance.

A MLLM-based method that infuses language instructions into grasp generation; & A new language-pose dataset, CapGrasp, featuring detailed caption of grasping poses.

A 4D motion dataset focusing on bimanual object manipulation tasks involved in complex daily activities with three-tiered task abstraction: Object Affordance, Primitive Task, and Complex Task.

A full-body motion dataset and pipeline for text-guided desktop object rearrangement with avatars.

Reconstructs colored neural implicit surfaces by decoupling geometry from view-dependent shading for accurate colored mesh extraction.

A category-level in-hand object reconstruction method with shape deformation priors and a synthetic COMIC dataset for diverse interactions.

A Transformer-based multi-view hand reconstruction method that uses point embeddings to fuse sparse camera observations into accurate hand meshes.

A richly textured MANO-derived articulated hand model with accessories, appearance variation, and benchmark-ready assets.

A large-scale hand-object interaction knowledge base with affordance annotations and a model for transferring grasps across objects.

An online exploration and synthesis framework that improves articulated hand-object pose estimation and generates dexterous grasps for novel objects.

Extends CPF into a unified spring-mass-style contact field that couples attraction, repulsion, and geometry constraints for robust hand-object reconstruction.

Models hand-object contact as a spring-mass-like potential field, enabling geometry-aware optimization for physically plausible interaction reconstruction.

Extends HybrIK to efficient whole-body mesh recovery by jointly modeling body, hands, and face.

A hybrid analytical-neural inverse kinematics method for fast and accurate 3D human pose and shape recovery.